We propose experiments to investigate population coding mechanisms in MT, a primate cortical visual area. There is now substantial evidence that MT neurons collectively encode visual motion, and that this information is accessed, or read out, in behaviors requiring motion detection. However, the nature of this code-the neural representation-is not understood. We will study two fundamental aspects of motion representation in MT. The first (Aim 1) is direction read-out, the mechanism by which many single-neuron activities are collectively interpreted as a single, overall direction. We will specifically determine whether the population of MT activities is interpreted in terms of the peak activity or in terms of the average activity of the cells. We will also determine whether neuron activities contain information about stimulus speed, or whether this information is constructed during the read-out process. Our second goal (Aim 2) is to understand segmentation mechanisms-how MT activities are grouped and assigned to a given moving object. We will test two basic hypotheses, the first stating that motion signals are perceptually bound, or assigned to the same visual object, when they generate a continuous pool of activity in MT. The second hypothesis is that motion signals are perceptually bound by virtue of synchronous firing in the MT population. We will test our ideas with monkeys trained to observe moving stimuli and report what they see. We will simultaneously measure the activities of MT neurons, and these neural data will be subsequently analyzed en masse with statistical models to find activity patterns that consistently foretell the animals' judgements. These will be the first studies to directly explore the link between distributed MT activity and the perception of visual motion.